Channel Modeling –ETI 085 Overview 9Department of Electrical and Information Technology fredrik...
Transcript of Channel Modeling –ETI 085 Overview 9Department of Electrical and Information Technology fredrik...
Channel Modeling – ETI 085gLecture no: 99
UWB Channel ModelingUWB Channel Modeling
Fredrik Tufvesson & Johan Kåredal, Department of Electrical and Information Technology
fredrik tufvesson@eit lth [email protected]
2009-11-26 Fredrik Tufvesson - ETI 085 1
Overview
( )?• What is Ultra-Wideband (UWB)?
• Why do we need UWB channel models?
• UWB channel modeling
• Standardized UWB channel models
• Summary
2009-11-26 Fredrik Tufvesson - ETI 085 2
What is Ultra-Wideband (UWB)?
• Transmitted power is spread over high bandwidth
• Definition:– Signals having
– and/orand/or
2009-11-26 Fredrik Tufvesson - ETI 085 3
Large Bandwidth Implications
f• High resistance to fading– Fine delay resolution; impulse response resolved into many delay-
binsbins– Fading within each delay-bin is smaller – Sum of all bins have even less fading
• Good ranging capability• Good wall and floor penetration (for some frequency
ranges)– Low-frequency components can go through material
2009-11-26 Fredrik Tufvesson - ETI 085 4
A Measured Impulse Response
BW = 7.5 GHz BW = 500 MHz
2009-11-26 Fredrik Tufvesson - ETI 085 5
Wireless Channel Bandwidth
Narrowband Wideband Ultra-wideband
frequency frequency frequency
delay delay delay
Increase in amplitude variation
Increase in delay variation
Two Possible UWB Techniques
• Pulse based UWB (impulse radio)– Transmission through ultra short time domain pulses in the
basebandbaseband– Evolution of the radar concept– Time hopping codes (Pulse Position Modulation)pp g ( )
• Multiband OFDM– OFDM-principle with frequency hopping in predefined subbands– Generation of UWB signals within carrier based systems
E i ll f hi h d t t t– Especially for high data rate systems
2009-11-26 Fredrik Tufvesson - ETI 085 7
Basic Principle
fUWB makes use of same spectrum as existing services:1. Information spread over wide spectrum; low power spectral
densitydensity2. Very low power
Small interference – looks like noise to other systems
802.11a (100MHz)
Part 15 LimitUWB (7.5 GHz)
Frequency3.1GHz
10.6GHz
5.725-5.825GHz
2009-11-26 Fredrik Tufvesson - ETI 085 8
Applications
• Personal area networksS ll– Small range
– Home networks (residential and office environments)– Consumer electronics
• Sensor networks– Lower data rate larger range (up to 300 m)
f– Typically for industrial environments• Other
– Military applications (frequency range < 1GHz )– Military applications (frequency range < 1GHz )– Geolocation– Through-wall radars
Viable candidate for several future applications!
2009-11-26 Fredrik Tufvesson - ETI 085 9
Frequency Regulations
• Regulations restrict frequency range that can be used• Measurements and models only practically useful in
that frequency range• FCC spectral mask:
-45
-40
m/M
Hz
-55
-50
n Le
vel i
n dB
-70
-65
-60
RP
Em
issi
on
1 2 3 4 5 6 7 8 910
-75
-70
Frequency in GHz
UW
B E
I
2009-11-26 Fredrik Tufvesson - ETI 085 10
Frequency in GHz
Frequency Regulations (cont’d)
SM
UWB
G
2009-11-26 Fredrik Tufvesson - ETI 085 11
A Fundamental Question
Q: Why do we need UWB Channel M d l ?Models?
A UWB h l f d llA: UWB channels are fundamentally different from narrowband channels.
Narrowband channel measurements and modeling cannot be reused!
2009-11-26 Fredrik Tufvesson - ETI 085 12
Narrowband vs. UWB Channel Models
• Assumptions about standard wireless channels:– “Narrowband” in the RF sense (bandwidth much smaller than carrier
frequency– WSSUS assumptionWSSUS assumption– Complex Gaussian fading (Rayleigh or Rice) in each delay tap
• Specialties of UWB channel:– Bandwidth comparable to carrier frequency– Different frequency components can “see” different reflection/
diffraction coefficients of obstacles– Few components per delay bin → central limit theorem (GaussianFew components per delay bin → central limit theorem (Gaussian
fading) not valid anymore
New channel models are needed!!
2009-11-26 Fredrik Tufvesson - ETI 085 13
New channel models are needed!!
Bandwidth Effect on Delay Tap Amplitude
Introduction
AmplitudeWideband: 0.1 GHzUltra-wideband: 7.5 GHz
Ultra-wideband is immune to multipath.p
Propagation Processes
Fundamental propagation processes:
F ti– Free space propagation– Reflection and transmission– DiffractionDiffraction– Diffuse scattering
All are frequency dependent!q y p
2009-11-26 Fredrik Tufvesson - ETI 085 15
Free-Space Propagation
Path gain of free-space propagation:
where the antenna gain is given by
Frequency dependent!
2009-11-26 Fredrik Tufvesson - ETI 085 16
Reflection and Transmission
• Dielectric properties of materials vary with frequency• Transmission (through two layered structure):
where the electrical length isgiven by
Frequency dependent!
2009-11-26 Fredrik Tufvesson - ETI 085 17
Diffraction
Diffraction from single screen:
Total electric field:
where andwhere and
Frequency dependent!
2009-11-26 Fredrik Tufvesson - ETI 085 18
Frequency dependent!
Scattering
Rough scattering according to Kirchoff theory:
roughf smooth exp −2 2 fc0h sin0
2
Frequency dependent!
2009-11-26 Fredrik Tufvesson - ETI 085 19
Frequency Dependency of UWB
Narrowband: 1 MHzWid b d 100 MHPropagation phenomina:
Free-space path-loss
Wideband: 100 MHzUltra-wideband: 7500 MHz
Dielectric layer transmission
Dielectric layer reflection
Ed diff tiEdge diffraction
Rough surface scattering
all propagation phenominas have a
frequency dependency.
UWB Channel ModelingUWB Channel Modeling
2009-11-26 Fredrik Tufvesson - ETI 085 21
Generic Channel Representation
• Tapped delay line model:pp y
• For UWB, each MPC show distortion:
where is the distortion functionwhere is the distortion function.
• Adjacent taps are influenced by a single physical MPC• Adjacent taps are influenced by a single physical MPC WSSUS assumption violated.
2009-11-26 Fredrik Tufvesson - ETI 085 22
Deterministic Modeling
S l M ll’ ti ith b d diti• Solve Maxwell’s equations with boundary conditions
“Exact” solutions• “Exact” solutions– Method of moments– Finite element method– Finite-difference time domain (FDTD)
• High frequency approximation– All waves modeled as rays that behave as in geometrical optics
t iray tracing– Refinements include approximation to diffraction, diffuse scattering,
etc.
2009-11-26 Fredrik Tufvesson - ETI 085 23
Principle of Ray Tracing
• Determine rays that can go from one TX position to one RX itiRX position
• Determine complex attenuation for all possible paths• Sum up contributions
2009-11-26 Fredrik Tufvesson - ETI 085 24
Deterministic Modeling for UWB
• Interaction processes now all depend on frequency and/or p p q y /direction
• Suggested solutions:– perform ray tracing at different frequencies, combine results– compute delay dispersion for each interaction process (possibly
different for different directions), concatenate
• Combine deterministic rays with diffuse clutter (statistically described)described)
2009-11-26 Fredrik Tufvesson - ETI 085 25
Statistical Channel Models
• Modeling of:g– Pathloss (total power)– Large-scale effects
• Shadowing• Delay dispersion (decay time constant)• Rice factorRice factor• Mean angle of arrival• “Parameters describing small-scale fading”
S ll l ff– Small-scale effects• Small-scale fading
2009-11-26 Fredrik Tufvesson - ETI 085 26
Modeling Path Gain
• Narrowband path gain:
• For UWB channel, define frequency-dependentth ipathgain:
• Simplified modeling:
2009-11-26 Fredrik Tufvesson - ETI 085 27
Modeling Path Gain (cont’d)
• Distance dependent path gain:p p g
– Path loss exponent varies from building to building → can be modeled as a random variable
• Frequency dependent path gain:
– κ varies between 0.8 and 1.4 (including antennas) and -1.4 and 1.5 (excluding antennas)
2009-11-26 Fredrik Tufvesson - ETI 085 28
Modeling Large-Scale Fading
• Defined as the variations of the local mean around the path pgain
• Commonly described as exhibiting a log-normal distribution
• Since large-scale fading is associated with diffraction and reflection effects, a frequency dependence would seem likely
• So far, measurements indicate no frequency dependence of shadowing variance
2009-11-26 Fredrik Tufvesson - ETI 085 29
Multi-Cluster Models
• How is a cluster determined?• Definition: components of cluster undergo same physical
processesp• Extraction from continuous measurements• Visual extraction from looks of (small-scale-averaged) ( g )
power delay profile• Fitting to measurement data
– Very sensitive to small changes• Better resolution when spatial information is taken into
account
2009-11-26 Fredrik Tufvesson - ETI 085 30
Saleh-Valenzuela Model
• Originally not for UWB [A.M. Saleh, R.A. Valenzuela, 1987]• MPCs arrive in clusters• Impulse responses given byg y
• Path interarrival times given by Poisson-distributed arrival process
• Different occurance rates for clusters (Λ) and rays (λ)
2009-11-26 Fredrik Tufvesson - ETI 085 31
Saleh-Valenzuela Model (cont’d)
Typical inter-cluster decay: 10-30 nsTypical intra-cluster decay: 1-60 ns
2009-11-26 Fredrik Tufvesson - ETI 085 32
Measured Power Delay Profile (LOS)
From 2m LOS measurement in factory hall:
-25-25
-35
[dB
] -35
[dB
]
-45
ved
pow
er
-45
ved
pow
er
-55Rec
eiv
-55Rec
eiv
0 20 40 60 80 100 120 140 160-65
τ [ns]0 20 40 60 80 100 120 140 160
-65
τ [ns]
2009-11-26 Fredrik Tufvesson - ETI 085 33
Generalizations
• Number of clusters as a random variable
• Cluster decay constants and arrival rates change with delay
l k Tl 0
• Ray arrival rates change with delay
• Cluster power varies due to shadowing
• Path interarrival times– Dense channel model - regularly spaced arrival times
S h l d l P i i l ti
2009-11-26 Fredrik Tufvesson - ETI 085 34
– Sparse channel model - Poisson arrival times
Measured Power Delay Profile (NLOS)
From NLOS measurement in factory hall:
-35
-40
[dB
]
-50
-45
ved
pow
er
-55
-50
Rec
eiv
0 20 40 60 80 100 120 140 160-60
τ [ns]
2009-11-26 Fredrik Tufvesson - ETI 085 35
τ [ s]
Modified Shape of Power Delay Profile
Can be modeled through a soft onset:
Power [dB]
t
2009-11-26 Fredrik Tufvesson - ETI 085 36
Small-Scale Fading Statistics
• Measurements report power within each bin being G di t ib t d lit d i N k iGamma-distributed, amplitude is m-Nakagami distributed:
h f t d l d d i blwhere m-factors are modeled as random variables
F di f d l bi i d l d l t d• Fading of delay bins is modeled as uncorrelated
Ph d l d if l di ib d• Phases modeled as uniformly distributed
2009-11-26 Fredrik Tufvesson - ETI 085 37
Other Small-Scale Distributions
• Lognormal: looks similar to Nakagami with large m
• Rayleigh: does usually not workRayleigh: does usually not work
• Rice:
can be converted to Nakagami (though slightly different tails):
2009-11-26 Fredrik Tufvesson - ETI 085 38
Standardized UWB Channel ModelsStandardized UWB Channel Models
2009-11-26 Fredrik Tufvesson - ETI 085 39
IEEE 802.15.3a
• For evaluation of model proposals, standard channel model established
• Theoretical model: is only basis, from which impulse response realizations are generated
• 4 radio environments, all indoor (residential and office):– LOS: 0-4m– NLOS: 0-4m– LOS: 4-10m– NLOS: heavy multipath
2009-11-26 Fredrik Tufvesson - ETI 085 40
Model Structure
• Saleh-Valenzuela model
• Multiple clusters, multiple paths within each cluster
• Small-scale fading is lognormal
• Superimposed lognormal cluster fading
• Pathloss model: free-space pathloss
2009-11-26 Fredrik Tufvesson - ETI 085 41
Channel ParametersTarget Channel Characteristics5
CM 11 CM 22 CM 33 CM 44
mτ [ns] (Mean excess delay) 5.05 10.38 14.18 m [ ] ( y)
rmsτ [ns] (rms delay spread) 5.28 8.03 14.28 25 NP10dB (number of paths within 10 dB of the strongest path)
35
NP (85%) (number of paths that capture 85% of channel energy)
24 36.1 61.54 capture 85% of channel energy) Model Parameters Λ [1/nsec] (cluster arrival rate) 0.0233 0.4 0.0667 0.0667 λ [1/nsec] (ray arrival rate) 2.5 0.5 2.1 2.1 Γ (cluster decay factor) 7.1 5.5 14.00 24.00
( d f ) 4 3 6 7 7 9 12γ (ray decay factor) 4.3 6.7 7.9 12
1σ [dB] (stand. dev. of cluster lognormal fading term in dB)
3.4 3.4 3.4 3.4
2σ [dB] (stand. dev. of ray lognormal fading term in dB)
3.4 3.4 3.4 3.4
xσ [dB] (stand. dev. of lognormal fading term for total multipath realizations in dB)
3 3 3 3
Model Characteristics5
τ 5 0 9 9 15 9 30 1mτ 5.0 9.9 15.9 30.1
rmsτ 5 8 15 25 NP10dB 12.5 15.3 24.9 41.2 NP (85%) 20.8 33.9 64.7 123.3 Channel energy mean [dB] -0.4 -0.5 0.0 0.3
2009-11-26 Fredrik Tufvesson - ETI 085 42
gy [ ]Channel energy std dev. [dB] 2.9 3.1 3.1 2.7
IEEE 802.15.4a (high-frequency model)
• More general:– Larger ranges– More environments– More general structure g
• Radio environments1 Indoor office– 1. Indoor office
– 2. Indoor residential– 3. Indoor industrial
4 O td– 4. Outdoor– 5. Agricultural areas/farms– 6. Body-worn devices
2009-11-26 Fredrik Tufvesson - ETI 085 43
Generic Model Structure
• PathlossSi l di t l– Simple distance power law
– No random variations of pathloss exponent– Lognormal shadowing for each clusterg g
• Delay dispersion– Saleh-Valenzuela model– Ray arrival times are mixed Poisson process– Cluster decay constants can increase with delay– Some environments have different shape of PDP (soft onset)p ( )
• Small-scale fading– Nakagami fading– m-factor independent of delay– First component of cluster can have larger m-factor
2009-11-26 Fredrik Tufvesson - ETI 085 44
Summary
• UWB is very promising area forh t k ( l t i )– home networks (consumer electronics)
– sensor networks– military applications
• Fundamental differences of UWB channels to narrowband channels– Propagation mechanisms processes are frequency dependent
Different small scale statistics of fading– Different small-scale statistics of fading– Sparse impulse responses occur
• Standard channel models will not work for the UWB channel• Standardized channel models:
– IEEE 802.15.3a model: were useful in the pastIEEE 802 15 4a model:– IEEE 802.15.4a model:
• Covers most interesting environments• Includes most relevant propagation effects
F hi h d l f
2009-11-26 Fredrik Tufvesson - ETI 085 45
• For high and low frequency range